Detecting and classifying the mechanics of cancer and non-cancer cells by machine learning algorithm

The global burden of cancer has increased in recent years, posing a major public health challenge. Generally, cancer cells are mutate from normal cells and have distinctive mechanical specifications. Despite significant progress in precision medicine, accurately distinguishing cancer cells remains challenging due to the inherent complexities in characterizing single-cell surface properties. In this study, we utilized atomic force microscopy (AFM) to obtain the mechanical properties of hepatic cells, hepatoma cells, gastric cells, and gastric cancer cells. Then, machine learning techniques were used to identify and classify the cancer and non-cancer cells through AFM-based mechanical characteristics. After computational training, the accuracy of classification and screening of four kinds of cells reached 98%, with an area under the receiver operating characteristic curve (AUC-ROC) value of 97.98%. Consequently, we successfully identified digestive system cancer cells and highlighted the valuable role of digital pathology in tumor cell diagnosis. This study provides an objective basis and a new research method for the diagnosis of hepatic cancer and gastric cancer, enriching the tumor cell detection scheme.